Acceptance sampling: calculated risk or clear advantage?

In an earlier post, we’ve briefly mentioned acceptance sampling as an arguable difference between Statistical Process Control (SPC) and Statistical Quality Control (SQC). Now, we’ll dive deeper into this topic, learn about the main approaches to acceptance sampling and how automation could help determine sampling frequency and acceptance limits.

Insights based on representative samples

To validate the quality of their products and ensure that the needs of their customers are met, organizations perform inspections, thorough analyses, and tests. However, depending on the product or on time-, volume-, and cost restrictions, sometimes it’s not feasible to test each unit of production. This is where acceptance sampling comes in.

By testing a sample that is representative, manufacturers can make inferences about the quality of the entire batch. These inspections can help determine whether products are within specifications, or if they need to be rejected. Acceptance sampling is then the number of tests passed.

First employed by the US military during World War II to control the quality of bullets, acceptance sampling has been widely used to reduce unacceptable lots. In industries with mass production and set procedures, this quality control method has proved to be very economical.

Number of defects versus measurable quality characteristics

Before they devise a sampling plan, manufacturers need to decide how many products to test through sampling and the number of defects acceptable within a batch. In other words – they need to establish the acceptable quality limit (AQL) or the worst tolerable process average that would still pass inspection. Once the testing criteria have been set, they can pick between two main approaches: acceptance sampling by attributes and acceptance sampling by variables. 

Acceptance sampling by attributes looks at the number of defective items or the number of defects per item sampled. For example, if cracks are the inspection criteria used by a manufacturer of lightbulbs, all products with a crack will be rejected, regardless of the size of the cracks or their number. Similarly, if a maximum of three defects is considered within specifications, counting two defects will not get the batch rejected. Through this type of test, manufacturers say ”yes” or ”no” to the quality of the inspected samples. 

In contrast to acceptance sampling by attributes, where no measurements are performed, acceptance sampling by variables accepts or rejects batches by measuring the samples’ quality characteristics. Going back to our previous example, the lightbulb manufacturer would probably measure the diameter of the bulb’s shell or the bulb’s length (or height). Through this type of inspection, manufacturers answer questions such as ”how good/how bad” are the inspected samples compared to the set specifications.

Approaches to acceptance sampling and how a QMS platform can help

Limitations and risks

While in some industries it’s advantageous to perform acceptance sampling, it’s also important to remember that this method offers insights into products that have already been produced. Unlike Statistical Process Control (SPC), which closely monitors the process and can identify areas of improvement, acceptance sampling won’t help reduce variation.

There are two main risks to not sampling the entire batch:

 

  1. Rejecting a good quality batch, which increases waste and is seen as a risk for the producer.
  2. Approving a bad quality batch, which runs the danger of releasing sub-standard products on the market and is seen as a risk for the consumer.

To prevent this from happening, manufacturers try to boost their quality control. In the times of Industry 4.0 and Smart Manufacturing, automation and robotics have transformed product inspections leading to faster and more accurate results, as it becomes more feasible to inspect every unit of production.

Acceptance sampling and AlisQI

In increasingly connected factories, data can be used to drive quality processes. Automating statistics with AlisQI translates into real-time information, out-of-spec alarms, and the ability to analyze trends. Seeing how our customers benefitted from automating analytics and SPC tools, we also use automation in acceptance sampling.

AlisQI measures the number of defects per product and per product group. With this newly gained insight which analyzes a period of up to 100 days, our system can easily define a sampling frequency and an acceptance limit in accordance with ISO and industry standards. This provides lab technicians with clear overviews and helps them devise weekly sampling plans which saves them a lot of time.

Conclusion

Through the fast insights it can provide, acceptance sampling has been considered an economical quality control method especially in industries with mass production. Using acceptance sampling by attributes, manufacturers approve or reject batches based on defective items or the number of defects per item. With acceptance sampling by variables, they measure the samples’ quality characteristics to determine if these are within specifications.

While acceptance sampling spares manufacturers the trouble of individual unit testing, there are clear limitations and risks for not inspecting the quality of the entire batch. To boost their quality management, manufacturers have looked at robotics and automation. These technological improvements can help with inspections – even with performing acceptance sampling. In addition, you can use AlisQI to help with defining sampling frequency and acceptance limits by product or product groups.

Should there still be any unanswered questions, do not hesitate to  contact us. In the meantime, visit our product section to learn more about the AlisQI platform and how it manages to boost quality management.

Want to see more? Request a demo and we will reach out to discuss your needs and AlisQI’s capabilities in full detail.